Pith. sign in

REVIEW 3 major objections 5 minor 57 references

InstanceSplat: Instance-Aware Feed-Forward 3D Gaussian Splatting for Scene Understanding

T0 review · 3 major / 5 minor · reviewed 2026-08-10 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗ Review JSON

Pith's one-line read InstanceSplat constructs, in a single forward pass, a 3D Gaussian representation that jointly encodes appearance, geometry, instance identity, and language-aligned semantics from pose-free multi-view images.

desk verdict A credible feed-forward 3DGS system that couples reconstruction with instance and semantic learning, with strong results but a few verification gaps—most notably the unquantified reliance on VGGT pseudo-geometry. read the letter →

arxiv 2608.07144 v1 pith:EVOO54EL submitted 2026-08-07 cs.CV

Minchao Jiang , Xiaoxuan Ma , Shunyu Jia , Haoru Wang , Zhang Liang , Wentao Zhu This is my paper · ORCID
classification cs.CV
keywords feed-forward3DGaussiansplattinginstance-awaresceneunderstandingpose-freemulti-viewreconstructioncross-viewinstanceconsistencyopen-vocabularysemanticsegmentationnovel-viewsynthesisgroundingjointlearning
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

The paper attempts to establish that instance identity can be built directly into a feed-forward 3D Gaussian representation rather than appended after reconstruction. From pose-free multi-view images, a single network outputs Gaussians that carry not only color and shape but also a compact instance embedding and a language-aligned semantic feature, so novel views, consistent object masks, and open-vocabulary labels all come from one representation. The authors claim that reconstruction, instance learning, and semantic learning reinforce each other when coupled through shared instance structure, yielding better rendering near object boundaries and more coherent object-level semantics. This matters because previous feed-forward methods are category-oriented, while instance-aware methods require slow per-scene optimization; if true, the approach brings instance-level 3D understanding to near-interactive speeds.

What carries the argument

The central object is an instance-aware Gaussian primitive, a 3D Gaussian that augments standard appearance and geometry attributes with an 8-dimensional instance embedding and a compact semantic feature. The mechanism that carries the argument is differentiable rendering of these feature fields from the shared Gaussians into the input views: because the same 3D primitives serve every view, image-space instance supervision is grounded in a common, renderable carrier, which yields cross-view-consistent identities. The supporting machinery is a set of three coupled loss families: a prototype-based contrastive objective (pull, push, and cross-view alignment) for instance grounding; an instance-boundary-weighted RGB loss for reconstruction; and a semantic-alignment loss whose similarity weights reweight the instance push term while instance regions pool semantic features for inference.

What would settle it

Swapping the pseudo-geometry distillation loss for ground-truth poses and depths on a benchmark that provides them, and observing whether instance segmentation mIoU rises substantially, would settle whether the method's ceiling is set by the pretrained geometry model's error.

Watch

Extended reading notes

Core claim

The central discovery is that instance identity is treated as an intrinsic attribute of each 3D Gaussian, not as a feature learned on top of a finished reconstruction. By differentiably rendering the instance embeddings from the shared Gaussians into supervised views and applying prototype-based contrastive losses with explicit cross-view alignment, the model learns a renderable, cross-view-consistent instance field. That grounded instance structure then acts as an interface: boundary-aware weighting focuses the reconstruction loss on sharp depth changes between objects, semantic similarity increases the separation pressure between confusing same-category instances, and the predicted instance regions aggregate language-aligned semantic features into stable object-level predictions. The result is a single feed-forward model that, on held-out indoor scenes and an unseen dataset, outperforms both feed-forward and per-scene-optimized baselines on novel-view synthesis, instance segmentation, and open-vocabulary understanding while taking roughly three seconds per scene.

Load-bearing premise

All geometry — camera poses, depths, and point maps — is learned from pseudo-labels produced by a pretrained geometry foundation model, so any systematic error in those pseudo-labels is inherited by the reconstructed Gaussians and by every instance or semantic feature built on them.

Editorial extensions

If this is right

  • A single feed-forward pass on unseen scenes yields renderable novel views, cross-view-consistent instance segmentation, and open-vocabulary semantic queries without per-scene optimization.
  • Instance boundaries sharpen the reconstructed depth and color near object contours, so reconstruction quality and instance quality improve together.
  • Cross-view instance tracking becomes more consistent because instance features are rendered from the same 3D Gaussians, and removing the cross-view alignment term measurably hurts tracking.
  • The representation supports object-level semantic aggregation, so open-vocabulary queries return coherent labels per instance rather than noisy per-pixel predictions.
  • On an unseen dataset the per-scene runtime is about three seconds, orders of magnitude faster than per-scene optimized instance-aware methods.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • An immediate extension the authors leave implicit is interactive 3D editing: because each object's identity is a Gaussian attribute, editing or deleting an instance could propagate consistently to all views, not just the edited one.
  • The heavy reliance on pseudo-geometry suggests that swapping in a geometry foundation model with better pose and depth accuracy could lift all downstream results; a controlled comparison would quantify that ceiling.
  • The instance embeddings, being renderable and language-aligned, could serve directly as a 3D prior for robotics tasks such as grasping or navigation, where object identity must persist across viewpoints.
  • The paper itself notes that the clustering readout becomes more expensive as views and Gaussians grow; replacing it with a learned grouping head would likely remove that scaling bottleneck.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

3 major / 5 minor

Summary. InstanceSplat proposes a unified feed-forward 3D Gaussian Splatting framework that, from pose-free multi-view images, constructs Gaussians jointly encoding appearance, geometry, instance identity, and language-aligned semantics in a single forward pass. The method introduces a 3D-Consistent Instance Grounding module with rendered-space contrastive losses and an Instance-Centric Coupling module comprising boundary-aware RGB supervision, semantic-guided hard-negative reweighting, and instance-grounded semantic aggregation. Experiments cover novel-view synthesis, instance segmentation, and open-vocabulary understanding on ScanNet, LERF, and the IGGT benchmark, claiming state-of-the-art performance, roughly 3-second inference, and cross-dataset generalization.

Significance. If fully substantiated, InstanceSplat would provide a practical unified representation for reconstruction and instance-level scene understanding, and the shared-Gaussian design is a sensible way to obtain cross-view-consistent instance features. Strengths of the manuscript include a coherent architecture, four ablations that individually test each coupling mechanism, a concrete training budget, and a zero-shot evaluation on LERF. However, the central state-of-the-art claim is undercut by incomplete reporting in Table 1, an ablation result that contradicts the claimed benefit of hard-negative reweighting, and the absence of any characterization of the model's sensitivity to the VGGT pseudo-geometry supervision on which all downstream geometry and features are built.

major comments (3)
  1. [Section 4.2, Table 1] The Ours row reports only mIoU and mAcc for the 2-view and 4-view settings, omitting PSNR, SSIM, and LPIPS, yet the text states that InstanceSplat achieves the best PSNR and LPIPS among methods reporting reconstruction metrics. Without these values, the NVS SOTA claim is not verifiable in the two primary settings; please report the full metrics and, ideally, error bars or a significance statement for all tables.
  2. [Section 4.2, Table 1 (4-view columns)] In the 4-view setting, the full model achieves mIoU 43.51 and mAcc 73.68, which are lower than the Ours w/o Hard Neg. row (mIoU 45.96, mAcc 73.77). This directly contradicts the Section 4.4 claim that semantic-guided hard-negative reweighting improves instance discrimination and weakens one of the central reciprocal-interaction claims; please explain this result or correct the ablation.
  3. [Section 3.1, Eq. (17) and Section 4.2] L_distill is the only term supervising camera poses and depths, and every downstream component (Gaussian centers, instance grounding, boundary-aware RGB loss, semantic aggregation) inherits errors from the VGGT pseudo-ground-truth geometry. The paper does not quantify this dependence, and the LERF experiment itself shows a drop from 8 to 16 views attributed to pose-accuracy degradation under domain shift. Please add an analysis of sensitivity to pseudo-label noise and, if feasible, compare against training or evaluation with ground-truth poses/depth on ScanNet/ScanNet++ to bound the generalization claim.
minor comments (5)
  1. [Table 1] The row for Ours appears merged with the preceding ablation row in the printed text; please format each row with clear separators so readers can distinguish the full model from the ablations.
  2. [Section 3.3, Eq. (11)] The boundary-aware RGB loss divides by the sum of boundary weights, which is undefined if all weights in a view are zero; please specify a numerical safety term or define the loss to be zero in that case.
  3. [Section 3.2, Eqs. (5) and (15)] The bracket notation [x]_+ is used in Eq. (5) without definition, while Eq. (15) defines [x]^1_0; please define both notations explicitly at first use.
  4. [Section 4.3, Table 3] The text refers to T-SR while the table caption specifies T-SR@0.5; please make the threshold explicit in the text and table body for reproducibility.
  5. [References] Several references are dated 2026 with arXiv identifiers (e.g., Ren et al., Yoo et al., Zheng et al.) and should be consistently marked as preprints or, if already published, updated with venue information.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: geometry, instance, and semantic supervision all come from external pretrained models and datasets, and the one same-author citation is not load-bearing.

full rationale

InstanceSplat's derivation chain is not circular. The geometry supervision is an external pseudo-label source, VGGT (Wang et al. 2025), used through L_distill in Eq. (17) to match predicted poses and depths; this is distillation from an independently trained model, not a quantity defined in terms of InstanceSplat's own outputs. Instance supervision comes from external refined GT masks (InsScene-15K from IGGT), and semantic supervision comes from the external LSeg teacher via Eq. (13). The inference-time HDBSCAN clustering and instance-grounded aggregation in Eq. (16) are post-processing compositions rather than fitted parameters renamed as predictions. The claimed reciprocal interactions are tested by ablations (Tables 1, 3, 4) that remove one module at a time, and the coupling terms in Eqs. (9)-(15) either use detached features or external teacher features, so no equation reduces a predicted quantity to its own input. The only same-author citation, VoteSplat (Jiang et al. 2025b), appears in Related Work as one example among several per-scene optimization methods and is not load-bearing. The skeptical concern that VGGT pseudo-geometry bounds cross-dataset generalization is a genuine robustness and correctness limitation, and the paper itself concedes a performance drop on LERF with 16 views due to domain shift and pose misalignment; however, this is not circularity because the pseudo-labels are external inputs rather than predictions of the paper.

Assumptions & free parameters 4 free parameters · 5 assumptions · 0 invented entities

The framework leans on several external pretrained components and hand-set objectives. The three most consequential are VGGT pseudo-geometry (which supervises all pose and depth predictions), LSeg teacher semantics (which defines the language alignment), and refined instance masks from InsScene-15K (which supervise the contrastive instance losses). The boundary-weight thresholds and loss weights are chosen by hand with no sensitivity study. No new physical or ontological entities are introduced; the instance embeddings and semantic features are learned attributes attached to existing Gaussian primitives.

free parameters (4)
  • margin hyperparameters delta_pull, delta_push, delta_cross = not reported
    Set by hand for the contrastive losses in Eqs. (5)-(7); central to instance embedding geometry but no sensitivity analysis is provided.
  • loss weights lambda_p, lambda_ins, lambda_sem, lambda_bd, lambda_pull, lambda_push, lambda_cross = 0.05, 0.01, 0.02, 0.02, 2, 1, 2
    Chosen in implementation; the balance between reconstruction, instance, and semantic objectives affects all reported metrics and is not varied in the ablations.
  • boundary weighting thresholds tau and T = tau=0.15, T=0.05
    Hand-set in Eq. (10); converts instance feature discontinuities into soft boundary weights that directly feed the boundary-aware RGB loss.
  • hard-negative reweighting parameters eta_max and tau_sem = eta_max and tau_sem not reported in main text
    Eq. (15) increases push weight for semantically similar instances; the actual values are not stated, so the behavior of the semantic-guided discrimination module is not fully reproducible.
assumptions (5)
  • domain assumption VGGT Geometry Foundation Model provides reliable pseudo-ground-truth camera poses and depth maps for training (Section 3.1, L_distill in Eq. (17)).
    If VGGT pseudo-labels contain systematic error, the predicted poses, depths, back-projected Gaussian centers, and all downstream instance and semantic grounding inherit the bias.
  • domain assumption LSeg teacher supplies reliable language-aligned semantic features for L_sem (Section 3.3, Eq. (13)).
    The semantic alignment and the semantic-guided hard-negative weighting both depend on the teacher's feature space being meaningful for open-vocabulary queries.
  • domain assumption Refined instance masks from InsScene-15K are accurate and cross-view consistent (Section 4.1).
    Equations (4)-(7) average rendered features within GT instance regions; errors in these masks directly corrupt the prototypes and the contrastive gradients.
  • ad hoc to paper The confidence-aware voxelization preserves instance boundaries and per-pixel feature information after merging (Section 3.1).
    The paper relies on this aggregation to reduce redundancy but provides no quantitative analysis of feature distortion or boundary loss introduced by the merging.
  • standard math 3DGS alpha-compositing is differentiable and adequate for rendering instance and semantic features (Section 3.2, Eq. (3)).
    Standard rendering model from Kerbl et al. 2023, applied here to 8-dim instance and semantic feature fields.

how reviews work

0 comments
Cite this review

Pith. "Pith review of InstanceSplat: Instance-Aware Feed-Forward 3D Gaussian Splatting for Scene Understanding." pith.science (2026). https://pith.science/paper/EVOO54EL

@misc{pith2026260807144,
  author       = {Pith},
  title        = {Pith review of: InstanceSplat: Instance-Aware Feed-Forward 3D Gaussian Splatting for Scene Understanding},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/EVOO54EL}},
  note         = {Machine review of arXiv:2608.07144}
}
read the original abstract

Feed-forward 3D Gaussian Splatting (3DGS) enables efficient and generalizable 3D reconstruction, but current feed-forward 3DGS methods for scene understanding remain largely category-oriented. In contrast, instance-aware 3DGS methods typically rely on per-scene optimization and often decouple reconstruction from instance and semantic learning, limiting reciprocal interactions among them. We present InstanceSplat, a unified feed-forward 3DGS framework for generalizable 3D reconstruction and instance-aware scene understanding from pose-free multi-view images. In a single forward pass, InstanceSplat constructs an instance-aware Gaussian representation that jointly encodes appearance, geometry, instance identity, and language-aligned semantics. Shared 3D Gaussians ground instance identities across views, producing renderable and cross-view-consistent instance features. To allow reconstruction and scene understanding to benefit from each other, we further design an instance-centric learning strategy that connects reconstruction, instance learning, and semantic learning through shared instance structure. Specifically, instance cues guide reconstruction, language-aligned semantics strengthen the discrimination of confusing same-category instances, and instance regions aggregate semantic evidence into coherent object-level predictions. Experiments on novel-view synthesis, instance segmentation, and open-vocabulary semantic understanding under varying input-view settings and on an unseen dataset demonstrate state-of-the-art performance, practical efficiency, and strong generalization.

Figures

Figures reproduced from arXiv: 2608.07144 by the authors.

Figure 1
Figure 1. Given pose-free multi-view images (a), Instance [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. InstanceSplat overview. Given pose-free multi-view images, a feed-forward backbone constructs an instance-aware [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figure 3
Figure 3. Visualization of instance-derived boundary [PITH_FULL_IMAGE:figures/full_fig_p004_3.png] view at source ↗
Figures from the paper (3 more)
Figure 4
Figure 4. Figure 4: Qualitative comparison with feed-forward 3DGS baselines on novel-view rendering and semantic segmentation. We [PITH_FULL_IMAGE:figures/full_fig_p006_4.png]
Figure 5
Figure 5. Figure 5: Visualization of the 3D feature fields after PCA [PITH_FULL_IMAGE:figures/full_fig_p006_5.png]
Figure 6
Figure 6. Figure 6: Qualitative analysis of instance representations. The left panel shows the effect of semantic-aware hard-negative [PITH_FULL_IMAGE:figures/full_fig_p007_6.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

57 extracted references · 19 canonical work pages

  1. [1]

    Proceedings of the Computer Vision and Pattern Recognition Conference , pages=

    Vggt: Visual geometry grounded transformer , author=. Proceedings of the Computer Vision and Pattern Recognition Conference , pages=

  2. [2]

    European conference on computer vision , pages=

    Grounding image matching in 3d with mast3r , author=. European conference on computer vision , pages=. 2024 , organization=

    2024

  3. [3]

    arXiv preprint arXiv:2408.13912 , year=

    Splatt3r: Zero-shot gaussian splatting from uncalibrated image pairs , author=. arXiv preprint arXiv:2408.13912 , year=

    arXiv

  4. [4]

    European Conference on Computer Vision , pages=

    Sparseneus: Fast generalizable neural surface reconstruction from sparse views , author=. European Conference on Computer Vision , pages=. 2022 , organization=

    work page 2022

  5. [5]

    Proceedings of the European Conference on Computer Vision (ECCV) , month =

    Yao, Yao and Luo, Zixin and Li, Shiwei and Fang, Tian and Quan, Long , title =. Proceedings of the European Conference on Computer Vision (ECCV) , month =

    work page

  6. [6]

    Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV) , month =

    Chen, Anpei and Xu, Zexiang and Zhao, Fuqiang and Zhang, Xiaoshuai and Xiang, Fanbo and Yu, Jingyi and Su, Hao , title =. Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV) , month =. 2021 , pages =

    2021

  7. [7]

    Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) , month =

    Yu, Alex and Ye, Vickie and Tancik, Matthew and Kanazawa, Angjoo , title =. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) , month =. 2021 , pages =

    work page 2021

  8. [8]

    , author=

    hdbscan: Hierarchical density based clustering. , author=. J. Open Source Softw. , volume=

Show all 57 references
  1. [9]

    Proceedings of the IEEE conference on computer vision and pattern recognition , pages=

    Scannet: Richly-annotated 3d reconstructions of indoor scenes , author=. Proceedings of the IEEE conference on computer vision and pattern recognition , pages=

  2. [10]

    Proceedings of the IEEE/CVF International Conference on Computer Vision , pages=

    Scannet++: A high-fidelity dataset of 3d indoor scenes , author=. Proceedings of the IEEE/CVF International Conference on Computer Vision , pages=

  3. [11]

    Foundations and Trends in Computer Graphics and Vision , volume=

    Multi-view stereo: A tutorial , author=. Foundations and Trends in Computer Graphics and Vision , volume=. 2015 , publisher=

    2015

  4. [12]

    and Frahm, Jan-Michael , title =

    Schonberger, Johannes L. and Frahm, Jan-Michael , title =. Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR) , month =

  5. [13]

    International Conference on Learning Representations , year=

    Language-driven Semantic Segmentation , author=. International Conference on Learning Representations , year=

  6. [14]

    European conference on computer vision , pages=

    Scaling open-vocabulary image segmentation with image-level labels , author=. European conference on computer vision , pages=. 2022 , organization=

    2022

  7. [15]

    arXiv preprint arXiv:2510.22706 , year=

    IGGT: Instance-Grounded Geometry Transformer for Semantic 3D Reconstruction , author=. arXiv preprint arXiv:2510.22706 , year=

  8. [16]

    Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV) , month =

    Ranftl, Ren\'e and Bochkovskiy, Alexey and Koltun, Vladlen , title =. Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV) , month =. 2021 , pages =

    2021

  9. [17]

    arXiv:2304.07193 , year=

    DINOv2: Learning Robust Visual Features without Supervision , author=. arXiv:2304.07193 , year=

    arXiv

  10. [18]

    Proceedings of the IEEE/CVF International Conference on Computer Vision , pages=

    In-place scene labelling and understanding with implicit scene representation , author=. Proceedings of the IEEE/CVF International Conference on Computer Vision , pages=

  11. [19]

    Panoptic Lifting for 3D Scene Understanding With Neural Fields , booktitle =

    Siddiqui, Yawar and Porzi, Lorenzo and Bul\`o, Samuel Rota and M\"uller, Norman and Nie. Panoptic Lifting for 3D Scene Understanding With Neural Fields , booktitle =. 2023 , pages =

    2023

  12. [20]

    European Conference on Computer Vision , pages=

    Click-gaussian: Interactive segmentation to any 3d gaussians , author=. European Conference on Computer Vision , pages=. 2024 , organization=

    2024

  13. [21]

    arXiv preprint arXiv:2410.24207 , year=

    No pose, no problem: Surprisingly simple 3d gaussian splats from sparse unposed images , author=. arXiv preprint arXiv:2410.24207 , year=

    arXiv

  14. [22]

    Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV) , month =

    Yang, Zesong and Yang, Bangbang and Dong, Wenqi and Cao, Chenxuan and Cui, Liyuan and Ma, Yuewen and Cui, Zhaopeng and Bao, Hujun , title =. Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV) , month =. 2025 , pages =

    2025

  15. [23]

    Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV) Workshops , month =

    Huang, Tianyu and Chen, Runnan and Hu, Dongting and Huang, Fengming and Gong, Mingming and Liu, Tongliang , title =. Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV) Workshops , month =. 2025 , pages =

    2025

  16. [24]

    Advances in neural information processing systems , volume=

    Large spatial model: End-to-end unposed images to semantic 3d , author=. Advances in neural information processing systems , volume=

  17. [25]

    European conference on computer vision , pages=

    Gaussian grouping: Segment and edit anything in 3d scenes , author=. European conference on computer vision , pages=. 2024 , organization=

    2024

  18. [26]

    arXiv preprint arXiv:2512.17541 , year=

    FLEG: Feed-Forward Language Embedded Gaussian Splatting from Any Views , author=. arXiv preprint arXiv:2512.17541 , year=

    arXiv

  19. [27]

    arXiv preprint arXiv:2508.03643 , year=

    Uni3r: Unified 3d reconstruction and semantic understanding via generalizable gaussian splatting from unposed multi-view images , author=. arXiv preprint arXiv:2508.03643 , year=

  20. [28]

    ACM Transactions on Graphics (TOG) , volume=

    Anysplat: Feed-forward 3d gaussian splatting from unconstrained views , author=. ACM Transactions on Graphics (TOG) , volume=. 2025 , publisher=

    2025

  21. [29]

    arXiv preprint arXiv:2410.06245 , year=

    Hisplat: Hierarchical 3d gaussian splatting for generalizable sparse-view reconstruction , author=. arXiv preprint arXiv:2410.06245 , year=

    arXiv

  22. [30]

    Proceedings of the Computer Vision and Pattern Recognition Conference , pages=

    Depthsplat: Connecting gaussian splatting and depth , author=. Proceedings of the Computer Vision and Pattern Recognition Conference , pages=

  23. [31]

    Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , pages=

    pixelsplat: 3d gaussian splats from image pairs for scalable generalizable 3d reconstruction , author=. Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , pages=

  24. [32]

    , author=

    3d gaussian splatting for real-time radiance field rendering. , author=. ACM Trans. Graph. , volume=

  25. [33]

    Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages=

    Langsplat: 3d language gaussian splatting , author=. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages=

  26. [34]

    Proceedings of the IEEE/CVF International Conference on Computer Vision , pages=

    Votesplat: Hough voting gaussian splatting for 3d scene understanding , author=. Proceedings of the IEEE/CVF International Conference on Computer Vision , pages=

  27. [35]

    arXiv preprint arXiv:2507.07136 , year=

    Langsplatv2: High-dimensional 3d language gaussian splatting with 450+ fps , author=. arXiv preprint arXiv:2507.07136 , year=

  28. [36]

    Proceedings of the Computer Vision and Pattern Recognition Conference , pages=

    Instancegaussian: Appearance-semantic joint gaussian representation for 3d instance-level perception , author=. Proceedings of the Computer Vision and Pattern Recognition Conference , pages=

  29. [37]

    Proceedings of the Computer Vision and Pattern Recognition Conference , pages=

    Panogs: Gaussian-based panoptic segmentation for 3d open vocabulary scene understanding , author=. Proceedings of the Computer Vision and Pattern Recognition Conference , pages=

  30. [38]

    1996 , publisher=

    Computer graphics: principles and practice , author=. 1996 , publisher=

    1996

  31. [39]

    Advances in Neural Information Processing Systems , volume=

    Opengaussian: Towards point-level 3d gaussian-based open vocabulary understanding , author=. Advances in Neural Information Processing Systems , volume=

  32. [40]

    Communications of the ACM , volume=

    Nerf: Representing scenes as neural radiance fields for view synthesis , author=. Communications of the ACM , volume=. 2021 , publisher=

    2021

  33. [41]

    Proceedings of the IEEE/CVF international conference on computer vision , pages=

    Lerf: Language embedded radiance fields , author=. Proceedings of the IEEE/CVF international conference on computer vision , pages=

  34. [42]

    arXiv preprint arXiv:2506.07491 , year=

    Spatiallm: Training large language models for structured indoor modeling , author=. arXiv preprint arXiv:2506.07491 , year=

  35. [43]

    European conference on computer vision , pages=

    Mvsplat: Efficient 3d gaussian splatting from sparse multi-view images , author=. European conference on computer vision , pages=. 2024 , organization=

    2024

  36. [44]

    arXiv preprint arXiv:2303.00905 , year=

    Open-world object manipulation using pre-trained vision-language models , author=. arXiv preprint arXiv:2303.00905 , year=

    arXiv

  37. [45]

    The International Journal of Robotics Research , volume=

    Foundations of spatial perception for robotics: Hierarchical representations and real-time systems , author=. The International Journal of Robotics Research , volume=. 2024 , publisher=

    2024

  38. [46]

    Proceedings of the IEEE conference on computer vision and pattern recognition , pages=

    Vision-and-language navigation: Interpreting visually-grounded navigation instructions in real environments , author=. Proceedings of the IEEE conference on computer vision and pattern recognition , pages=

  39. [47]

    Proceedings of the Computer Vision and Pattern Recognition Conference , pages=

    Thinking in space: How multimodal large language models see, remember, and recall spaces , author=. Proceedings of the Computer Vision and Pattern Recognition Conference , pages=

  40. [48]

    arXiv preprint arXiv:2602.12159 , year=

    3DGSNav: Enhancing Vision-Language Model Reasoning for Object Navigation via Active 3D Gaussian Splatting , author=. arXiv preprint arXiv:2602.12159 , year=

  41. [49]

    FirstName Alpher , title =

  42. [50]

    Journal of Foo , volume = 13, number = 1, pages =

    FirstName Alpher and FirstName Fotheringham-Smythe , title =. Journal of Foo , volume = 13, number = 1, pages =

  43. [51]

    Journal of Foo , volume = 14, number = 1, pages =

    FirstName Alpher and FirstName Fotheringham-Smythe and FirstName Gamow , title =. Journal of Foo , volume = 14, number = 1, pages =

  44. [52]

    FirstName Alpher and FirstName Gamow , title =

  45. [53]

    Computer Vision -- ECCV 2022 , year =

    2022

  46. [54]

    An, Honggyu and Jung, Jaewoo and Kim, Mungyeom and Kim, Chaehyun and Jeon, Minkyeong and Han, Jisang and Fukuda, Kazumi and Narihira, Takuya and Ko, Hyuna and Kim, Junsu and others , journal=

  47. [55]

    arXiv preprint arXiv:2604.15239 , year=

    TokenGS: Decoupling 3D Gaussian Prediction from Pixels with Learnable Tokens , author=. arXiv preprint arXiv:2604.15239 , year=

    arXiv

  48. [56]

    arXiv preprint arXiv:2606.29513 , year=

    Scenes as Objects, Not Primitives: Instance-Structured 3D Tokenization from Unposed Views , author=. arXiv preprint arXiv:2606.29513 , year=

    arXiv

  49. [57]

    Proceedings of 1st international conference on image processing , volume=

    Two deterministic half-quadratic regularization algorithms for computed imaging , author=. Proceedings of 1st international conference on image processing , volume=. 1994 , organization=

    1994

Pith tools

Reviewed August 10, 2026 · model on record in the stance chip above.